Add serve_bread HumanEgo checkpoint (inference latest.pt + full latest_full.pt) with migrated config.json (object_centric + camera_frame + max_ict)
Browse files- serve_bread/config.json +138 -0
- serve_bread/dataset_stats.json +14 -0
- serve_bread/latest.pt +3 -0
- serve_bread/latest_full.pt +3 -0
serve_bread/config.json
ADDED
|
@@ -0,0 +1,138 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"out_dir": "./runs/serve_bread/Baselines/01_HumanEgo",
|
| 3 |
+
"MPS_PATHS_TRAIN": [
|
| 4 |
+
"./data/serve_bread/mps_serve_bread_001_vrs",
|
| 5 |
+
"./data/serve_bread/mps_serve_bread_002_vrs",
|
| 6 |
+
"./data/serve_bread/mps_serve_bread_003_vrs",
|
| 7 |
+
"./data/serve_bread/mps_serve_bread_004_vrs",
|
| 8 |
+
"./data/serve_bread/mps_serve_bread_005_vrs",
|
| 9 |
+
"./data/serve_bread/mps_serve_bread_006_vrs",
|
| 10 |
+
"./data/serve_bread/mps_serve_bread_007_vrs",
|
| 11 |
+
"./data/serve_bread/mps_serve_bread_008_vrs",
|
| 12 |
+
"./data/serve_bread/mps_serve_bread_009_vrs",
|
| 13 |
+
"./data/serve_bread/mps_serve_bread_010_vrs",
|
| 14 |
+
"./data/serve_bread/mps_serve_bread_011_vrs",
|
| 15 |
+
"./data/serve_bread/mps_serve_bread_012_vrs",
|
| 16 |
+
"./data/serve_bread/mps_serve_bread_013_vrs",
|
| 17 |
+
"./data/serve_bread/mps_serve_bread_014_vrs",
|
| 18 |
+
"./data/serve_bread/mps_serve_bread_015_vrs",
|
| 19 |
+
"./data/serve_bread/mps_serve_bread_016_vrs",
|
| 20 |
+
"./data/serve_bread/mps_serve_bread_017_vrs",
|
| 21 |
+
"./data/serve_bread/mps_serve_bread_018_vrs",
|
| 22 |
+
"./data/serve_bread/mps_serve_bread_019_vrs",
|
| 23 |
+
"./data/serve_bread/mps_serve_bread_020_vrs",
|
| 24 |
+
"./data/serve_bread/mps_serve_bread_021_vrs",
|
| 25 |
+
"./data/serve_bread/mps_serve_bread_022_vrs",
|
| 26 |
+
"./data/serve_bread/mps_serve_bread_023_vrs",
|
| 27 |
+
"./data/serve_bread/mps_serve_bread_024_vrs",
|
| 28 |
+
"./data/serve_bread/mps_serve_bread_025_vrs",
|
| 29 |
+
"./data/serve_bread/mps_serve_bread_026_vrs",
|
| 30 |
+
"./data/serve_bread/mps_serve_bread_027_vrs",
|
| 31 |
+
"./data/serve_bread/mps_serve_bread_028_vrs",
|
| 32 |
+
"./data/serve_bread/mps_serve_bread_029_vrs",
|
| 33 |
+
"./data/serve_bread/mps_serve_bread_030_vrs",
|
| 34 |
+
"./data/serve_bread/mps_serve_bread_031_vrs",
|
| 35 |
+
"./data/serve_bread/mps_serve_bread_032_vrs",
|
| 36 |
+
"./data/serve_bread/mps_serve_bread_033_vrs",
|
| 37 |
+
"./data/serve_bread/mps_serve_bread_034_vrs",
|
| 38 |
+
"./data/serve_bread/mps_serve_bread_035_vrs",
|
| 39 |
+
"./data/serve_bread/mps_serve_bread_036_vrs",
|
| 40 |
+
"./data/serve_bread/mps_serve_bread_037_vrs",
|
| 41 |
+
"./data/serve_bread/mps_serve_bread_038_vrs",
|
| 42 |
+
"./data/serve_bread/mps_serve_bread_039_vrs",
|
| 43 |
+
"./data/serve_bread/mps_serve_bread_040_vrs",
|
| 44 |
+
"./data/serve_bread/mps_serve_bread_041_vrs",
|
| 45 |
+
"./data/serve_bread/mps_serve_bread_042_vrs",
|
| 46 |
+
"./data/serve_bread/mps_serve_bread_043_vrs",
|
| 47 |
+
"./data/serve_bread/mps_serve_bread_044_vrs",
|
| 48 |
+
"./data/serve_bread/mps_serve_bread_045_vrs",
|
| 49 |
+
"./data/serve_bread/mps_serve_bread_046_vrs",
|
| 50 |
+
"./data/serve_bread/mps_serve_bread_047_vrs",
|
| 51 |
+
"./data/serve_bread/mps_serve_bread_048_vrs",
|
| 52 |
+
"./data/serve_bread/mps_serve_bread_049_vrs",
|
| 53 |
+
"./data/serve_bread/mps_serve_bread_050_vrs",
|
| 54 |
+
"./data/serve_bread/mps_serve_bread_051_vrs",
|
| 55 |
+
"./data/serve_bread/mps_serve_bread_052_vrs",
|
| 56 |
+
"./data/serve_bread/mps_serve_bread_053_vrs",
|
| 57 |
+
"./data/serve_bread/mps_serve_bread_054_vrs",
|
| 58 |
+
"./data/serve_bread/mps_serve_bread_055_vrs",
|
| 59 |
+
"./data/serve_bread/mps_serve_bread_056_vrs",
|
| 60 |
+
"./data/serve_bread/mps_serve_bread_057_vrs",
|
| 61 |
+
"./data/serve_bread/mps_serve_bread_058_vrs",
|
| 62 |
+
"./data/serve_bread/mps_serve_bread_059_vrs",
|
| 63 |
+
"./data/serve_bread/mps_serve_bread_060_vrs"
|
| 64 |
+
],
|
| 65 |
+
"MPS_PATHS_EVAL": [
|
| 66 |
+
"./data/serve_bread/mps_serve_bread_000_vrs"
|
| 67 |
+
],
|
| 68 |
+
"data_num": 100,
|
| 69 |
+
"task": "serve_bread",
|
| 70 |
+
"img_name": "rgb_WoArm_WArmObjKpts.png",
|
| 71 |
+
"centric_mode": "object_centric",
|
| 72 |
+
"frame_mode": "camera_frame",
|
| 73 |
+
"action_mode": "absolute",
|
| 74 |
+
"use_pcd_features": false,
|
| 75 |
+
"use_aux_obj_dynamics": false,
|
| 76 |
+
"use_aux_visual_foresight": false,
|
| 77 |
+
"use_aux_temporal_contrastive": false,
|
| 78 |
+
"use_region_attn": true,
|
| 79 |
+
"use_ot_cfm": false,
|
| 80 |
+
"enable_augmentation": true,
|
| 81 |
+
"enable_aug_img": true,
|
| 82 |
+
"enable_aug_rrc": true,
|
| 83 |
+
"enable_aug_target_jittering": true,
|
| 84 |
+
"enable_aug_cutout": true,
|
| 85 |
+
"enable_aug_temporal_stride": false,
|
| 86 |
+
"enable_aug_interpolation": true,
|
| 87 |
+
"use_amp": true,
|
| 88 |
+
"use_lr_schedule": true,
|
| 89 |
+
"warmup_steps": 200,
|
| 90 |
+
"min_lr_ratio": 0.05,
|
| 91 |
+
"image_size": [
|
| 92 |
+
240,
|
| 93 |
+
320
|
| 94 |
+
],
|
| 95 |
+
"pred_horizon": 50,
|
| 96 |
+
"single_hand": true,
|
| 97 |
+
"single_hand_side": "right",
|
| 98 |
+
"max_ict": 8,
|
| 99 |
+
"batch_size": 32,
|
| 100 |
+
"epochs": 400,
|
| 101 |
+
"lr": 0.0001,
|
| 102 |
+
"weight_decay": 0.01,
|
| 103 |
+
"num_workers": 8,
|
| 104 |
+
"grad_clip": 1.0,
|
| 105 |
+
"use_ema": true,
|
| 106 |
+
"ema_decay": 0.999,
|
| 107 |
+
"w_flow": 1.0,
|
| 108 |
+
"w_pos": 5.0,
|
| 109 |
+
"w_rot": 1.0,
|
| 110 |
+
"w_g": 10.0,
|
| 111 |
+
"w_done": 0.2,
|
| 112 |
+
"w_foresight": 20.0,
|
| 113 |
+
"w_contrastive": 0.1,
|
| 114 |
+
"use_pre_norm": true,
|
| 115 |
+
"use_ctx_norm": true,
|
| 116 |
+
"use_done_in_flow": false,
|
| 117 |
+
"use_legacy_image_loading": true,
|
| 118 |
+
"use_legacy_rng": true,
|
| 119 |
+
"use_state_noise": false,
|
| 120 |
+
"state_pos_noise_std": 0.001,
|
| 121 |
+
"state_rot_noise_deg": 1.0,
|
| 122 |
+
"state_grasp_noise_std": 0.0,
|
| 123 |
+
"num_inference_steps": 20,
|
| 124 |
+
"hand_tracking_method": "aria_mps",
|
| 125 |
+
"model_horizon_weighting": "uniform",
|
| 126 |
+
"model_horizon_beta": 0.0,
|
| 127 |
+
"patch_size": 16,
|
| 128 |
+
"vision_embed_dim": 384,
|
| 129 |
+
"num_decoder_layers": 6,
|
| 130 |
+
"num_heads": 8,
|
| 131 |
+
"mlp_ratio": 4.0,
|
| 132 |
+
"dropout": 0.05,
|
| 133 |
+
"eval_every": 1,
|
| 134 |
+
"vis_eval_every": 100,
|
| 135 |
+
"make_video": true,
|
| 136 |
+
"device": "cuda",
|
| 137 |
+
"seed": 7
|
| 138 |
+
}
|
serve_bread/dataset_stats.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"pos": {
|
| 3 |
+
"mean": [
|
| 4 |
+
0.05616205185651779,
|
| 5 |
+
-0.011201439425349236,
|
| 6 |
+
0.6174575090408325
|
| 7 |
+
],
|
| 8 |
+
"std": [
|
| 9 |
+
0.1598508358001709,
|
| 10 |
+
0.1118927076458931,
|
| 11 |
+
0.13916686177253723
|
| 12 |
+
]
|
| 13 |
+
}
|
| 14 |
+
}
|
serve_bread/latest.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7a90fee7dad1feb2a27c8ee6580f9811f42cdcc3836c77eafee17ab27be01385
|
| 3 |
+
size 60736310
|
serve_bread/latest_full.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5d2dbb39bfc32cfb0d49b33cf925dbce637266671cb7aca48067b03f1576e6a7
|
| 3 |
+
size 182254111
|